Comparative Analysis of Machine Learning Models for Auto Insurance Fraud Detection

Abstract

Auto insurance fraud is a significant problem in the insurance industry, with losses estimated to be in the billions of dollars annually. Insurance fraud can take many forms, including staged accidents, false injury claims, and exaggerated damage claims. The impact of auto insurance fraud is significant for both insurance companies and policyholders. Insurance fraud results in higher premiums for policyholders, as insurance companies pass on the costs of fraudulent claims to their customers. Fraudulent claims can also result in delays in claims processing and payouts, as insurance companies must investigate claims to determine their validity. Goal is to create a solution that analyzes auto insurance claim data to identify cases of fraud claims. Using machine learning algorithms, models are built to classify these claims. Also, This study compares the average accuracy, precision, recall, and other characteristics of all classification based machine learning methods. A machine learning model is created for fraudulent transaction validation using the scikit-learn Python Library.